Integrative bioinformatics analysis and elastic network modeling elucidate the role of cellular senescence in meningioma recurrence

Abstract

Background

Cellular senescence is intimately tied to tumorigenesis and progression, yet its exploration in meningiomas remains inadequate. In this study, we aim to unravel the role of cellular senescence-associated genes (CSA-genes) in meningioma recurrence and identify potential diagnostic markers and therapeutic targets.

Methods

We analyzed GSE136661 and GSE173825 datasets to identify CSA-signature genes through differential expression analysis, weighted gene co-expression network analysis, protein-protein interaction network construction, and elastic net regression modeling. Functional enrichment, immune cell infiltration using CIBERSORT, and transcription factor prediction were performed. Potential drugs were screened using Enrichr database.

Results

A total of 1827 differentially expressed genes (DEGs) were identified, among which 48 were cell senescence-associated differentially expressed genes (CSA-DEGs). Four key CSA-signature genes (CDK1, FOXM1, MYBL2, and BIRC5) were discovered by integrating elastic net regression and network algorithms. The elastic net model demonstrated strong classification performance with an area under the curve (AUC) of 0.816 in distinguishing recurrent meningiomas. Recurrent tumors exhibited significant immune heterogeneity, including increased neutrophils and M0 macrophages (p = 0.007), and CSA-genes were significantly correlated with immune infiltration and checkpoint molecules such as VSIR (p < 0.05). Transcription factor E2F1 was identified as a potential regulator of CSA-signature genes. Drug screening highlighted Dasatinib and Rapamycin as promising candidates with notable anti-meningioma potential.

Conclusion

Our findings highlight crucial genes and pathways in meningioma recurrence, introducing novel therapeutic candidates. These findings pave new avenues for further elucidating meningioma recurrence mechanisms and developing innovative treatments.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-025-03611-y.

Introduction

Meningiomas, the most prevalent primary tumors of the central nervous system, constitute 53% of non-malignant central nervous system (CNS) neoplasms, with an incidence rate of 7.86 cases per 100,000 individuals annually [1–3]. Although most meningiomas are benign, tumors in surgically challenging locations (e.g., skull base meningiomas) or those of higher grades often exhibit brain invasion and a high propensity for recurrence, even after multiple rounds of surgery, chemotherapy, and radiotherapy. Studies report recurrence rates of approximately 90% for these subtypes [4]. Recurrence of meningiomas adversely affects patient survival, with 10-year overall survival rates of 81.4% for non-malignant meningiomas and 57.1% for malignant ones, particularly dismal at 0% for grade III meningiomas [2]. Furthermore, a striking 20% of World Health Organization (WHO) grade I meningiomas recur following complete resection [5], underscoring the existence of underlying biological mechanisms yet to be elucidated.

The surgical resection or radiotherapy of recurrent meningiomas remains a formidable challenge, prompting the frequent consideration of systemic therapies. However, meningiomas have historically been understudied diseases, with evidence for systemic treatments generally scarce [6, 7]. Several compounds have been explored in small prospective studies, but while preliminary evidence suggests antitumor activity in patients with recurrent meningiomas, subsequent trials have failed to confirm significant clinical benefits [7, 8]. Thus, unraveling the underlying biological mechanisms and investigating novel targeted therapeutic strategies hold paramount importance. Delving deeper into these avenues has the potential to yield transformative treatments and improve outcomes for patients with recurrent meningiomas.

In recent years, cellular senescence strategies have garnered considerable attention. However, the relationship between cellular senescence and tumorigenesis is intricate. Initially, cellular senescence was deemed an effective antitumor mechanism [9, 10], capable of inhibiting tumorigenesis by inducing cell cycle arrest and promoting apoptosis, thereby restricting the proliferation, invasion, and metastasis of tumor cells [11]. Yet, increasing research has revealed that senescent cells (SNCs) may exert unexpected adverse effects on antitumor therapies [12]. Senescent cells secrete a variety of factors, a phenomenon ubiquitously known as the senescence-associated secretory phenotype (SASP) [13]. The accumulation of SASP in the tumor microenvironment promotes the expression of matrix metalloproteinases (MMPs) by activating multiple signaling pathways. MMPs degrade the extracellular matrix (ECM), releasing epithelial transforming factors and growth factors, which accelerate further tumor growth and epithelial-mesenchymal transition, thereby fostering tumor cell proliferation and invasion [14]. Additionally, senescence escape in tumor cells is a significant contributor to the failure of antitumor therapies and tumor recurrence [15], whereas the elimination of senescent cells delays tumorigenesis [16], both suggesting a protumorigenic role for tumor cell senescence. Meningiomas are diseases whose incidence significantly increases with age [17, 18]. As the basic unit of organismal aging, the role of cellular senescence in the initiation, progression, and particularly recurrence of meningiomas remains unknown. Based on this, we hypothesize that cellular senescence plays a crucial role in meningioma recurrence. This study will comprehensively utilize bioinformatics and elastic networks to investigate recurrent meningiomas in public databases, aiming to elucidate the role of cellular senescence in meningioma recurrence and provide insights for targeted meningioma therapies.

Materials and methods

Bulk-RNA sequencing data sources

We commence by outlining the study’s workflow in Fig. 1. The bulk-RNA sequencing transcriptome data for meningiomas were sourced from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo). Our search strategy encompassed the following criteria: (1) a keyword search for “Meningioma”; (2) selection of “Expression profiling by array” under the Study type option; (3) samples derived from Homo sapiens; and (4) datasets encompassing both primary and recurrent meningioma samples. Specifically, the mRNA sequencing for dataset GSE136661 was performed on the GPL20301 platform, comprising 15 recurrent meningioma samples and 145 primary meningioma samples. Meanwhile, dataset GSE173825, based on the GPL16791 platform, contained 5 recurrent meningioma samples and 3 primary meningioma samples. For this study, GSE136661 served as the discovery set, while GSE173825 functioned as the validation set. For the GSE136661 and GSE173825 datasets, we employed identical data processing workflows, including independent preprocessing and normalization, recurrent) as a covariate, to ensure that critical biological signals were preserved while eliminating technical batch effects. The Cellular Senescence Gene Database (http://csgene.bioinfo-minzhao.org) provided a List of 503 genes associated with cellular senescence.

Fig. 1.

Technical roadmap of this study

Differential expression analysis

Our initial analysis of the bulk-RNA sequencing datasets was conducted using R software (version 4.3.2). Prior to data analysis, rigorous data cleaning procedures were implemented, including normalization using the “NormalizeBetweenArrays” function and subsequent log2 transformation. Differentially expressed genes (DEGs) were identified utilizing the “Limma” package [19]. Initially, a design matrix was constructed to reflect the relationships between experimental and control groups. Subsequently, the lmFit function was employed to fit a linear model, and the eBayes function was applied for empirical Bayes moderation to calculate statistical significance. We established selection criteria of |log2 fold change| >0.5 and adjusted P-value < 0.05 to identify DEGs. Finally, the results were visually represented through volcano plots and heatmaps.

Identification of cell senescence-associated differentially expressed genes

To pinpoint differentially expressed genes that are intimately linked to cellular senescence, we employed a Venn diagram approach. By comparing the identified DEGs with a known set of genes associated with cellular senescence, we filtered out those DEGs that overlapped with the senescence-related gene list. This process allowed us to precisely identify genes that play pivotal roles in the cellular senescence process.

Gene enrichment analysis

To gain deeper insights into the differential genes, we utilized the “clusterProfiler” package (version 4.10) [20] for ID conversion and subsequent enrichment analysis. This encompassed Gene Ontology Biological Process (GO_BP) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis [21]. Entries with an FDR < 0.05 were considered significantly enriched, providing valuable insights into the biological processes and pathways associated with the differential genes.

Construction of protein-protein interaction (PPI) network

To construct a protein-protein interaction (PPI) network encompassing the cell senescence-associated differentially expressed genes (CSA-DEGs), we leveraged the STRING database version 12 (http://string-db.org). During network construction, we set the “minimum required interaction score” to 0.4 to ensure the reliability of interactions and opted to hide disconnected nodes for a simplified network structure. This approach yielded interaction scores among the CSA-DEGs. Subsequently, these data were imported into Cytoscape software version 3.9.1 [22] for further visualization. To identify key genes within the network, we employed the “cytoHubba” plugin [23] in Cytoscape to score all genes. Specifically, we ranked genes using the Betweenness and Degree algorithms, ultimately selecting the top 10 genes ranked by both methods as our key genes.

Weighted gene co-expression network analysis (WGCNA)

WGCNA is a systems biology approach used to identify gene modules closely related to specific phenotypes and explore the interrelationships among these genes. In this study, we employed the R package “WGCNA” [24] for our analysis. Initially, the GSE136661 expression data was cleaned and processed, with missing values handled and the top 15,000 genes with the highest median absolute deviation selected as the data source for analysis. Next, hierarchical clustering of samples was performed to assess similarity among samples and identify outliers for exclusion.

Selecting an optimal soft-thresholding power, a crucial step in constructing a scale-free network, was done to ensure the network’s stability. We set the networkType parameter to “unsigned” and RsquaredCut to 0.9. Subsequently, a co-expression network was constructed using the one-step method, with corType set to “bicor”, minModuleSize set to 100, and mergeCutHeight set to 0.25. Gene modules were identified using the dynamic tree cut method, and the correlation and significance of each module’s eigengene with the phenotype were calculated.

Ultimately, this process enabled us to identify the gene modules most significantly associated with specific traits and the genes within these modules.

Elastic net regression

Elastic Net regression is a widely-used regularization technique in high-dimensional data analysis that combines the strengths of Lasso regression (L1 regularization) and Ridge regression (L2 regularization) [25]. Its core idea lies in addressing multicollinearity and variable selection issues in high-dimensional data by incorporating both regularization terms [25, 26]. In this study, we adopted elastic net regression methods and oversampling techniques to improve the balance of samples. To address sample imbalance, we applied the Synthetic Minority Over-sampling Technique (SMOTE), with parameters perc.over = 200 and perc.under = 150, increasing minority class samples to 200% of their original number while under-sampling the majority class to 150%. We used k = 5 nearest neighbors to generate synthetic samples, ensuring a near 1:1 ratio between positive (recurrent) and negative (non-recurrent) classes. Optimal regularization parameters (alpha and lambda) were selected through cross-validation to identify key genes associated with specific groupings.

First, we divided the data into 5-fold cross-validation sets using the caret package to ensure the robustness of model evaluation. For each fold, we conducted a grid search over different alpha values ranging from 0 to 1 with a step size of 0.1. The cv.glmnet function was used to perform binomial logistic regression cross-validation, selecting the model with the lowest classification error on the validation set. Finally, the model parameters with the highest accuracy across all folds were chosen to fit the final Elastic Net model on the entire dataset.

By analyzing the model coefficients, we identified genes with non-zero coefficients as the key genes. To evaluate the classification performance of the model, we utilized the “pROC” package [27] to generate ROC curves and calculate the AUC values. This comprehensive approach demonstrated the effectiveness of Elastic Net in gene selection and classification tasks.

Validation and evaluation of CSA-signature genes

To validate and assess the CSA-signature genes, we first obtain the intersection of key module genes from WGCNA, Elastic Net results, MCC (Mutual Coherence Criterion), and Degree algorithm results using a Venn diagram. This intersection represents the CSA-signature genes that are consistently identified by multiple methods.

Immune infiltration analysis

To evaluate the landscape of immune infiltration, we used the ESTIMATE algorithm to assess tumor purity in the samples. The data were adjusted by incorporating tumor purity as a covariate; regression analysis was performed to correct the proportions of immune cell infiltration, thereby minimizing the potential confounding effect of tumor purity. Subsequently, we first conducted a quantitative analysis of the immune cell composition within the test dataset using the “CIBERSORT” algorithm [28]. We then delved into the correlation between the expression levels of CSA-signature genes and the abundance of various immune cell populations. Specifically, we calculated the Pearson correlation coefficients and their corresponding p-values between gene expression levels and immune cell densities. Correlations with p-values less than 0.05 were considered statistically significant. Finally, we assessed the relationship between the CSA-signature genes and immune checkpoints using Pearson analysis. To visually represent these correlations, we generated a heatmap using the “ggplot2” package.

Transcription factor screening and evaluation

To identify key transcription factors (TFs) that regulate the cellular senescence-associated signature genes (CSA-signature genes), we utilized the transcription factor prediction module of the NetworkAnalyst 3.0 web platform (https://www.networkanalyst.ca). This platform offers robust analytical tools for delving into regulatory networks embedded within gene expression data [29]. Initially, we imported the CSA-signature genes into the NetworkAnalyst platform to obtain a list of potential transcription factors. Subsequently, we evaluated the expression levels of these predicted TFs across different groups within the two datasets.

Identifying candidate drugs

To pinpoint potential candidate drugs, we submitted the CSA-signature genes to the Enrichr website (https://www.amp.pharm.mssm.edu/Enrichr/) and downloaded the information on target drugs associated with these genes. Subsequently, we ranked the candidate drugs based on a comprehensive score, arranged from highest to lowest. This comprehensive score reflects the degree of association between each small molecule drug and the studied genes. Drugs with a significant level (P-value < 0.05) and a high comprehensive score were deemed prominent candidates for further consideration.

Statistical analysis

Statistical analyses were performed using R (version 4.3.2). To estimate the differences in expression between two groups, the Wilcoxon rank-sum test was employed. A P-value of less than 0.05 was considered statistically significant.

Results

Cell senescence-associated DEGs and functional enrichment analysis

In the dataset GSE136661, mRNA sequencing data from both primary tumor and recurrent tumor groups successfully passed quality control measures, revealing no significant outliers and substantial gene expression differences between the two groups (Fig. 2A–C). Consequently, all samples were included in subsequent analyses. Differential expression analysis yielded 1827 differentially expressed genes (DEGs) (|log2 fold change| >0.5 and adjusted P-value < 0.05) (Fig. 2D), of which 847 were downregulated and 980 were upregulated. Further comparison with a curated set of 503 cell senescence-related genes identified 48 cell senescence-associated differentially expressed genes (CSA-DEGs) (Fig. 2E), comprising 35 upregulated and 13 downregulated genes.

Fig. 2.

Differentially Expressed Genes (DEGs) Related to Cellular Senescence and Functional Enrichment Analysis. A–C Sample data quality validation from dataset GSE136661 shows no significant outliers, with clear differentiation between the two groups. D Volcano plot depicting the DEGs identified in dataset GSE136661. E Venn diagram illustrating the intersection of DEGs from dataset GSE136661 and cellular senescence-associated genes, yielding the CSA-DEGs

Enrichment analysis of the CSA-DEGs revealed significant associations with key biological processes and pathways. GO analysis highlighted cell cycle regulation and DNA replication (P < 0.001, FDR < 0.05) (Supplementary Material 2: Figure S1B), while KEGG identified the P53 signaling pathway and cellular senescence as top enriched terms (P < 0.01) (Supplementary Material 2: Figure S1C). These findings implicate senescence-related mechanisms in meningioma recurrence.

WGCNA key module genes

During the WGCNA, we first ensured the quality of the data, confirming the absence of significant outlier samples (Fig. 3A). Subsequently, A soft-threshold power of 12 was chosen to achieve a scale-free topology fit index (R²) > 0.9, ensuring network robustness. The moderate correlation (r = 0.42) aligns with typical co-expression network strengths in transcriptomic studies, resulting in the identification of four distinct modules (Fig. 3B, C). Correlation analysis revealed that the MEgreen module exhibited the highest correlation with the phenotype of interest (r = 0.42, p = 2e–08, Fig. 3D).

Fig. 3.

Identification of Key Module Genes Using WGCNA. A Dendrogram indicating no significant outliers in the data. B Soft-thresholding power analysis suggests an optimal soft-thresholding power of 12. C, D WGCNA yields four modules, with MEgreen showing the strongest correlation with the trait of interest. E Bubble plot displaying the results of enrichment analysis for genes within the MEgreen module

Within the MEgreen module, we identified 251 genes, 846 genes in the MEblue module, 385 genes in the MEbrown module, 1017 genes in the METurquoise module, and 221 genes in the MEyellow module (detailed in Supplementary Material 1), which were further subjected to enrichment analysis. The results of Biological Process (BP) and KEGG pathway enrichment analyses highlighted terms such as cellular senescence, cell cycle, apoptosis, and the P53 signaling pathway (Fig. 3E, note: original figure reference adjusted for context). These findings underscore the intimate relationship between cellular senescence and the recurrence phenotype in meningiomas, suggesting that the genes within the MEgreen module may play crucial roles in mediating this process.

PPI network construction and functional enrichment analysis

Initially, we intersected the genes within the MEgreen module with the CSA-DEGs, resulting in the identification of 25 CSA-Hub genes (Fig. 4A). Subsequently, utilizing the STRING database, we constructed a PPI (Protein–Protein Interaction) network for the CSA-DEGs. This network comprises 25 nodes interconnected by 246 edges, demonstrating a statistically significant PPI enrichment with a p-value of < 1.0e–16.

Fig. 4.

Construction of PPI Network and Functional Enrichment Analysis. A Venn diagram depicting the intersection between genes in the MEgreen module and CSA-DEGs, resulting in 25 CSA-Hub genes. B PPI network of CSA-related genes generated using the Degree algorithm. C, D Bubble plots displaying the results of GO and KEGG enrichment analysis for CSA-related genes

To evaluate the importance of individual genes within this network, we employed the ‘cytoHubba’ plugin to score all genes based on commonly used algorithms: Degree and Betweenness. This analysis identified key genes and their corresponding scores (Fig. 4B), with detailed results provided in Supplementary Material 1.

Following this, we conducted GO functional enrichment and KEGG pathway enrichment analyses to delve into the biological pathways implicated in meningioma recurrence. The GO biological process enrichment analysis uncovered significant enrichments in crucial cellular processes such as cell cycle transition, DNA damage-responsive signal transduction, and DNA replication (Fig. 4C).

Meanwhile, the KEGG pathway enrichment analysis highlighted essential pathways including cellular senescence, apoptosis, and the P53 signaling pathway (Fig. 4D). These findings collectively emphasize the pivotal role of cellular senescence in the mechanisms governing meningioma recurrence, providing valuable insights into the underlying biological processes and pathways.

Elastic net model analysis

To further identify key genes related to cellular senescence in the recurrence mechanism of meningiomas, we analyzed a dataset comprising 1,827 DEGs. Firstly, we conducted a Principal Component Analysis (PCA) on the expression matrix of CSA-Hub genes, resulting in a clear differentiation between the two groups of samples, 19 genes with non-zero coefficients (Fig. 5A). Subsequently, through the establishment of an Elastic Net model, we identified the top 20 genes ranked by their importance (Fig. 5B). This model demonstrated robust performance in distinguishing between primary and recurrent meningiomas, with an AUC of 0.816 (95% CI: 0.78–0.85), indicating high diagnostic accuracy (Fig. 5C). Next, we intersected the top 10 genes ranked by the Degree and Betweenness algorithms with those ranked by the Elastic Net model’s importance, ultimately pinpointing four CSA-signature genes: CDK1, FOXM1, MYBL2, and BIRC5 (Fig. 5D). Further GO enrichment analysis of these CSA signature genes revealed crucial biological processes such as cell cycle transition and DNA damage-responsive signal transduction (Supplementary Material 2: Figure S2A). Additionally, KEGG pathway enrichment analysis illuminated the associations of these genes with essential pathways including cellular senescence, apoptosis, and the P53 signaling pathway (Supplementary Material 2: Figure S2B). Furthermore, Wikipathway enrichment analysis indicated their involvement in pathways such as DNA IR Damage and Cell Response Via ATR, CKAP4 signaling, and IL-24 signaling (Supplementary Material 2: Figure S2C). Lastly, through MSigDB Hallmark enrichment analysis, we discovered that these genes are primarily associated with processes like G2/M checkpoint, E2F Targets, and Mitotic Spindle (Supplementary Material 2: Figure S2D). These findings not only reinforce the pivotal role of CSA-signature genes in meningioma recurrence but also suggest that they might play a significant part in this process by influencing cellular proliferation and DNA repair mechanisms.

Fig. 5.

Elastic Net Selection of Key Genes and Enrichment Analysis of CSA-Signature Genes. A PCA plot demonstrating the ability of CSA-Hub genes to distinguish between primary and recurrent meningiomas. B Ranking of gene importance based on the elastic net model. C ROC curve illustrating the excellent discriminative ability of the elastic net model. D Venn diagram showcasing the identification of four CSA-signature genes

Immunoinfiltration analysis

Given that cellular senescence modifies the tumor microenvironment and is closely associated with immune responses. We employed the CIBERSORT algorithm to quantitatively assess the infiltration of immune cells within tumor samples. Initially, the CIBERSORT analysis revealed significant heterogeneity in immune cell infiltration among meningioma samples (Fig. 6A), with notable differences in the distribution of various immune cell types between primary and recurrent tumor groups (Fig. 6B). Specifically, in the recurrent tumor group, there was a marked increase in the proportions of neutrophils (p = 0.04) and M0 macrophages (p = 0.007), accompanied by a significant decrease in the proportions of eosinophils (p = 0.03) and naive B cells (p = 0.03). Subsequently, we examined the correlation between the CSA-signature genes and the proportions of immune cells. The results indicated a close association between these CSA-signature genes and the aforementioned changes in immune cell proportions (Fig. 6C). Lastly, we discovered a significant correlation between all CSA-signature genes and the V-domain Ig suppressor of T cell activation receptor (VSIR) gene (p = 0.05) (Fig. 6D). Through this analysis, we have uncovered distinct differences in immune cell infiltration between primary and recurrent meningiomas, and identified the potential roles of CSA-signature genes within the tumor immune microenvironment.

Fig. 6.

Immune Infiltration Analysis in Meningiomas. A Heatmap depicting significant heterogeneity in immune cell infiltration between primary and recurrent meningioma samples. B Bar chart displaying differences in immune cell proportions between primary and recurrent meningioma groups. C Heatmap showing the correlation between CSA-signature genes and immune cell types. D Heatmap illustrating the correlation between CSA-signature genes and immune checkpoints

Validation and evaluation of CSA-signature genes

Initially, we conducted a differential analysis of the CSA-signature genes within the dataset GSE136661. The results demonstrated that the mRNA expressions of CDK1, FOXM1, MYBL2, and BIRC5 were significantly higher in recurrent meningiomas compared to primary meningiomas, with all P < 0.001 (Fig. 7A). Furthermore, these four CSA-signature genes exhibited excellent diagnostic performance, with AUC values of 0.813, 0.849, 0.872, and 0.840, respectively (Fig. 7B). These findings were validated in the dataset GSE173825 (Figs. 7C, D), further underscoring the reliability and generalizability of our results. Additionally, the data processing details for the dataset GSE173825 are provided in Supplementary Material 2: Figure S3.

Fig. 7.

Validation of CSA-Signature Genes and Exploration of Candidate Transcription Factors. A Bar chart displaying the differential expression analysis of CSA-signature genes in dataset GSE136661. B ROC curve demonstrating the excellent diagnostic performance of CSA-signature genes in dataset GSE136661. C, D Differential expression analysis and diagnostic performance of CSA-signature genes in dataset GSE173825. E, F Validation of candidate transcription factors in datasets GSE136661 and GSE173825

Exploration and evaluation of transcription factors

To gain further insights into the regulatory mechanisms of the CSA-signature genes, we conducted an exploration and evaluation of relevant transcription factors. Utilizing the TRRUST and JASPAR databases, we identified candidate transcription factors that potentially regulate the expression of CDK1, FOXM1, MYBL2, and BIRC5 (Supplementary Material 2: Figure S4). Among these, KLF4 and E2F1 emerged as potential regulators. Subsequently, we performed differential analysis of the mRNA expression levels of KLF4 and E2F1 in the datasets GSE136661 and GSE173825. Our findings revealed that the expression of E2F1 was significantly different between primary and recurrent meningiomas, whereas no significant difference was observed in KLF4 expression (Fig. 7E, F). This suggests that E2F1 may play a crucial role in modulating the expression of the CSA-signature genes and potentially contributing to the differences observed in the immune microenvironment between primary and recurrent meningiomas.

Identification of candidate drugs

To determine potential therapeutic agents for treating recurrent meningiomas, we employed the Enrichr website and inputted the CSA-signature genes for drug screening. Leveraging the enrichment analysis from the DsigDB database, we generated a List of the top 10 candidate drugs based on their Combined Score, as outlined in Table 1. These candidate drugs encompass a diverse range of chemical structures and mechanisms of action, offering a broad spectrum of therapeutic options. These drugs have demonstrated varying degrees of anti-meningioma activity in vitro and in vivo studies, with Dasatinib and Rapamycin emerging as particularly promising candidates due to their notable potential for further development into clinical therapeutic agents.

Table 1.

Top 10 candidate drugs selected through DsigDB enrichment analysis

Term	Adjusted P-value	Combined score	Genes
Phytoestrogens CTD 00007437	1.04E–08	1935924.864	CDK1;BIRC5;MYBL2;FOXM1
LUCANTHONE CTD 00006227	2.22E–06	1440253.305	CDK1;BIRC5;MYBL2;FOXM1
Dasatinib CTD 00004330	4.64E–05	1150486.005	CDK1;BIRC5;MYBL2;FOXM1
troglitazone CTD 00002415	9.87E–05	1061016.686	CDK1;BIRC5;MYBL2;FOXM1
Enterolactone CTD 00001393	3.92E–04	921506.0298	CDK1;BIRC5;MYBL2;FOXM1
rapamycin CTD 00007350	4.36E–04	2869.860506	CDK1;BIRC5;FOXM1
roscovitine CTD 00003426	5.34E–04	8747.566939	CDK1;BIRC5
7-Hydroxystaurosporine CTD 00002331	5.34E–04	8021.859107	CDK1;BIRC5
Fulvestrant CTD 00002740	5.34E–04	2142.505433	CDK1;BIRC5;FOXM1
testosterone CTD 00006844	5.34E–04	832973.8493	CDK1;BIRC5;MYBL2;FOXM1

Discussion

In this study, we have not only unveiled the potential role of cellular senescence in meningioma recurrence but also elucidated the significance of CSA-signature genes through an integrative approach encompassing various analytical methods such as WGCNA, elastic net, and PPI-related algorithms. Furthermore, we have systematically delved into their functions and mechanisms from perspectives including immune infiltration, functional enrichment, and transcriptional factor regulation, thereby offering crucial biological insights and clinical implications. Several transcriptome-based studies have identified key genes and signaling pathways involved in meningioma recurrence, such as Cao et al., who found that CXCL2 and CXCL8 are associated with immune infiltration and recurrence [30], and Zhang et al., who revealed that FOXC2 promotes tumor progression by regulating ADAM12 to activate the JAK1/STAT3/VEGFA pathway [31]. In contrast, our study is the first to focus on cellular senescence-related genes, proposing a regulatory network involving key genes such as CDK1 and FOXM1 and the transcription factor E2F1, thereby enriching the molecular understanding of meningioma recurrence and providing new avenues for targeted therapy.

Cellular senescence emerges as a crucial factor in the recurrence mechanisms of meningioma. Previous studies have demonstrated its dual role in tumor suppression and cancer progression [30, 32]. In this study, we observed significant alterations in the expression of CSA-DEGs in recurrent meningiomas, particularly the upregulation of CSA-signature genes, suggesting their potential key role in tumor recurrence. To further elucidate the functions of these genes, we utilized WGCNA to identify critical modules associated with meningioma recurrence, notably the MEgreen module. Genes within this module were enriched in biological processes related to cellular senescence, cell cycle regulation, apoptosis, and the P53 signaling pathway, further reinforcing the significance of cellular senescence in meningioma recurrence. Subsequently, we conducted GO and KEGG enrichment analyses on CSA-DEGs, CSA-related Hub genes, and CSA-signature genes, revealing their primary involvement in crucial biological processes such as the cell cycle, DNA replication, and the P53 signaling pathway. This indicates that CSA-signature genes may be involved in the process of tumor recurrence, which needs to be verified by subsequent functional experiments. Additionally, Wikipathway enrichment analysis provided a novel perspective, suggesting that CSA-signature genes may be implicated in the radiotherapy resistance mechanisms of recurrent meningiomas, particularly through the ATR pathway [33]. These findings offer a systematic view, enhancing our understanding of the vital mechanisms of cellular senescence in meningioma recurrence.

Meningiomas are not constrained by the blood-brain barrier, allowing surrounding immune cells to infiltrate these tumors. The composition of immune infiltration within the tumor microenvironment is intimately linked to tumor progression [34]. CIBERSORT analysis has revealed remarkable heterogeneity in immune cell infiltration patterns among meningioma samples, with notable increases in the proportions of neutrophils and M0 macrophages in recurrent meningiomas. These alterations may be associated with an immunosuppressive state within the tumor microenvironment [35, 36]. Moreover, CSA-signature genes exhibit a close correlation with changes in immune cell proportions, particularly displaying significant positive correlations with M0 macrophages and Treg cells. This suggests that these genes may play a role in meningioma recurrence by modulating the immune microenvironment, consistent with previous studies [37]. Intriguingly, CSA-signature genes display negative correlations with immune checkpoint molecules, notably VSIR (V-domain Ig suppressor of T cell activation receptor), a crucial immune checkpoint that plays a pivotal role in regulating immune responses [38]. This finding implies that CSA-signature genes may influence tumor immune evasion mechanisms by modulating the expression of immune checkpoint molecules.

Through this analysis, we have not only uncovered significant differences in immune cell infiltration between primary and recurrent meningiomas but also identified the potential role of CSA-signature genes in the tumor immune microenvironment. These discoveries offer new avenues for further research into meningioma immunotherapy and provide a theoretical foundation for developing therapeutic strategies targeting specific immune cell types and immune checkpoint molecules.

The identification of four key CSA-signature genes—CDK1, FOXM1, MYBL2, and BIRC5—through an elastic net model is a significant finding, as these genes demonstrate exceptional diagnostic performance in distinguishing between primary and recurrent meningiomas, validated in external datasets. Prior research underscores the significance of these genes in meningioma biology: CDK1 is notably associated with meningioma recurrence and biological aggressiveness [39], FOXM1 is linked to the enhancement of meningioma malignant features [40, 41], MYBL2 mutations are correlated with high-grade meningiomas [42], and a NanoString targeted gene expression panel study on meningioma progression highlights the crucial roles of FOXM1, MYBL2, and BIRC5 [43]. Radiotherapy, chemotherapy, surgical treatment, and alterations in the post-operative tumor microenvironment may lead to DNA damage, oxidative stress, or cell cycle arrest in tumor cells, thereby inducing senescence in meningioma cells. This cellular senescence phenomenon is not entirely irreversible, particularly in treatment-induced senescent cells, where some senescent cancer cells can escape cell cycle arrest. Tumor cells that undergo senescence escape can acquire stemness and re-enter a new cell cycle to replenish the tumor, leading to cancer recurrence. Further enrichment analysis reveals that these genes are primarily involved in crucial biological processes such as cell cycle transition and DNA damage response, suggesting they may contribute to meningioma recurrence by regulating cell proliferation and DNA repair. To gain deeper insights into their regulatory mechanisms, we explored potential transcription factors for these CSA-signature genes and identified E2F1 as a candidate. E2F1 is a key transcription factor involved in cell proliferation, differentiation, and apoptosis [44] and is currently considered a potential therapeutic target in various cancers, including colorectal, breast, and gastric cancers [44–46]. By analyzing the co-expression relationship between E2F1 and the CSA-signature genes in the two datasets, we ensure the reliability of its regulatory role, but it has not been directly verified in meningiomas. This discovery not only enhances our understanding of the molecular mechanisms underlying meningioma recurrence but also opens up new avenues for developing targeted therapies that can potentially disrupt the oncogenic pathways driven by these genes and their regulators.

Utilizing the Enrichr database, we have identified several potential therapeutic drugs. To date, Dasatinib and Rapamycin are the most extensively studied drugs. Dasatinib, a multi-targeted tyrosine kinase inhibitor, has demonstrated anti-proliferative and anti-metastatic activities in various tumors. In meningioma, previous studies have suggested that Dasatinib can slow tumor growth by inhibiting the Eph receptor tyrosine kinase (EPH RTK) [47]. Our study reveals associations between CSA genes and Dasatinib targets, further supporting the potential of Dasatinib as a therapeutic option for meningioma. Rapamycin, an mTOR inhibitor, has been widely investigated for targeted therapy in central nervous system tumors. Preclinical studies have shown that Rapamycin can induce proliferation inhibition and cell cycle arrest in meningioma cells [48]. In our study, CSA genes are enriched in cell cycle and DNA damage response pathways, suggesting that Rapamycin may exert its therapeutic effects on recurrent meningioma by modulating cellular senescence pathways. Nevertheless, clinical data on these drugs for meningioma remain limited, and further in vivo and clinical studies are needed to evaluate their safety and efficacy. Additionally, the other candidate drugs identified in this study have not been previously reported, providing novel clues for drug development.

Study limitation and future prospects

Despite the insights gained from this study into the potential role of cell aging-related genes in meningioma recurrence, there are still some limitations. Firstly, this study is based on the bioinformatics analysis of public transcriptome data, lacking the functional verification of cell and animal models, and unable to determine the specific pathogenic mechanism of CSA gene and its direct role in meningioma recurrence. Secondly, due to limitations in the dataset, we were unable to obtain specific clinical parameters for prognostic analysis. However, we believe that our study still offers a preliminary methodological foundation for identifying the biological characteristics of recurrent meningiomas. Our research emphasizes potential biological mechanisms underlying the upregulation of certain genes in recurrent meningiomas, which can guide future studies. Moreover, it should be noted that the test dataset in this study exhibits some heterogeneity, and thus the results should not be over-extrapolated. Future studies should incorporate functional assays, such as gene knockdown or overexpression, animal models of meningioma recurrence, and clinical cohort analyses to systematically verify the biological functions and clinical relevance of CSA genes. Additionally, integrative multi-omics approaches—including genomics, epigenomics, and single-cell sequencing—will be instrumental in fully elucidating the molecular mechanisms underlying meningioma recurrence. Although we have preliminarily validated candidate drugs, further in vivo and in vitro experiments are required to assess their safety and efficacy.

Additionally, this study performed pre-selection of cellular senescence-associated genes (CSA-genes) prior to enrichment analysis may introduce bias. To minimize potential bias, we also performed GO and KEGG enrichment analyses on all DEGs (1827 genes) from the GSE136661 dataset. However, in order to enhance the sensitivity of Our mechanistic analysis and prevent key signals from being diluted by noise in whole-genome analysis, as well as to maintain the conciseness of the manuscript, these results are presented in Supplementary Material 2: Figure S1A.

Conclusion

In conclusion, through systematic analysis, this study has revealed the pivotal role of cellular aging in the recurrence mechanism of meningioma and provided new avenues for future therapeutic strategies. These findings not only deepen our understanding of the recurrence mechanisms of meningioma but also offer potential biomarkers and drug targets for individualized treatment.

Abbreviations

CSA-genes

Cellular senescence-associated genes

GEO

Gene Expression Omnibus

DEG

Differentially expressed gene

GO

Gene ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

WGCNA

Weighted gene co-expression network analysis

PPI

Protein–protein interaction

Author contributions

JHH and CHL analyzed the data and wrote this manuscript. YC, and YBK assisted in analyzing the data and revising the manuscript. JHH, YC critically read and edited the manuscript.

Funding

None.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. GSE43292: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE136661 GSE163154: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc= GSE173825 Cell Senescence Database: http://csgene.bioinfo-minzhao.org.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.